Project Objectives

Physically-based rendering (PBR) requires an accurate reproduction of the physical behavior of light and its interaction with the surface materials, which is formalized by the rendering equation. The typical approach to solve the illumination integral is by resorting to Classical Monte Carlo (CMC), thanks to its simplicity and ease of implementation. However, CMC requires a large number of samples to obtain an accurate estimate, even when coupled with variance reduction techniques. To address this issue several methods have been proposed in the literature, among which Bayesian Monte Carlo (BMC) and other Gaussian Process-based approaches with promising results. Nevertheless, the use of BMC in CG is still in an incipient phase and its application to more evolved and widely used rendering algorithms remains cumbersome. This research project proposes to investigate and develop efficient solutions to generalize the application of this very promising technique to PBR.

Candidate Qualifications

Candidates must hold a PhD in computer science, physics, mathematics, or in a related discipline, and have a strong publication record in their respective field. A solid mathematical background is required, as well as good programming skills in C++, Python and/or Matlab. Experience in Rendering techniques and/or Machine Learning is highly valued. Experience with Monte Carlo methods is also a plus.

Work Environment

The work will be developed within the context of a Research Consolidation grant (09/2023 to 08/2025) awarded to Dr. Ricardo Marques, and will be carried on within the Department of Mathematics and Informatics of the University of Barcelona, a multidisciplinary and extremely dynamic research environment. The selected candidate will have the opportunity collaborate with a PhD Student on the topic of Bayesian Monte Carlo for Photo-Realistic Rendering currently supervised by Dr. Ricardo Marques at University Pompeu Fabra, as well as to benefit from the national and international research collaborations of Dr. Ricardo Marques. For further inquiries please contact Dr. Ricardo Marques (ricardo.marques@ub.edu).